In [54]:
import os
import subprocess
import donuts.deconv.utils as du
import donuts.deconv.ncx as dn
import nibabel as nib
import numpy as np
import numpy.random as npr
import scipy.optimize as spo
import numpy.linalg as nla
from operator import add
from pyspark.mllib.clustering import KMeans
import matplotlib.pyplot as plt
In [39]:
from pyspark.mllib.clustering import KMeans
In [2]:
os.chdir('/root/data')
subprocess.check_output('aws s3 cp s3://rawpredator/8907_7/b0_corrected.nii.gz 8907_7_b0_corrected.nii.gz', shell=True)
Out[2]:
In [42]:
coil_data1 = nib.load('/root/data/8907_7_b0_corrected.nii.gz').get_data()
coil_data1 = coil_data1[:, :, :, 2:]
In [43]:
np.shape(coil_data1)
Out[43]:
In [9]:
def standardize(x):
return (x - np.mean(x))/np.std(x)
def sdot(a, b):
return np.squeeze(np.dot(a, b))
def par(st):
return np.array([float(v) for v in str(st).split(' ')])
def depar(x):
return ' '.join(str(xx) for xx in x)
def arr_red(x, y):
return np.vstack([x, y])
def array_to_rdd(a, parts = 48, rm = False, fn = 0):
if fn == 0:
fn = 'temp' + str(npr.randint(0, 999999)) + '.txt'
os.chdir('/root/ephemeral-hdfs/bin')
np.savetxt(fn, a, fmt = '%.9e')
os.system('./hadoop fs -put ' + fn + ' ' + fn)
rawrdd = sc.textFile(fn, parts).map(par).cache()
rawrdd.count()
if rm:
os.system('rm ' + fn)
return rawrdd
def rdd_to_array(rawrdd, parts = 48, rm = False, fn = 0):
if fn == 0:
fn = 'temp' + str(npr.randint(0, 999999)) + '.txt'
os.chdir('/root/ephemeral-hdfs/bin')
rawrdd.map(depar).saveAsTextFile(fn)
os.system('./hadoop fs -getmerge ' + fn + ' ' + fn)
a = np.loadtxt(fn)
# cleanup
os.system('rm ' + fn)
os.system('./hadoop fs -rmr ' + fn)
return a
def rank1_approx(a):
u, s, v = nla.svd(a, False)
return s[0] * np.outer(u[:, 0], v[0, :])
In [44]:
def demean(x):
return x - np.mean(x)
coil_dm1 = np.apply_along_axis(demean, 3, coil_data1)
In [6]:
np.mean(coil_dm1[20, 20, 20, :])
Out[6]:
In [7]:
np.shape(coil_dm1)
Out[7]:
In [45]:
w_inds = range(44)
vmin_2 = -1.0
vmax_2 = 1.0
z_inds = [30, 40, 50, 60]
figsize_ = (3, 3)
figsize_2 = (len(z_inds) * figsize_[0], len(w_inds) * figsize_[1])
f_sv, axarr = plt.subplots(len(w_inds), len(z_inds), sharex=True, sharey=True, figsize = figsize_2)
for i in range(len(w_inds)):
w_ind = w_inds[i]
for j in range(len(z_inds)):
z_ind = z_inds[j]
axarr[i, j].imshow(coil_dm1[:, :, z_ind, w_ind], cmap='bone', vmin = vmin_2, vmax = vmax_2)
